From 07772fcf6fca44217439154aa37e4854dd5aef34 Mon Sep 17 00:00:00 2001 From: Ralf Jung Date: Wed, 29 Apr 2020 14:56:40 +0200 Subject: [PATCH] expand comment in memory.rs with extra soundness concerns --- src/librustc_mir/interpret/memory.rs | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/src/librustc_mir/interpret/memory.rs b/src/librustc_mir/interpret/memory.rs index 0d0ed465c1cc..8560594b583d 100644 --- a/src/librustc_mir/interpret/memory.rs +++ b/src/librustc_mir/interpret/memory.rs @@ -453,7 +453,9 @@ impl<'mir, 'tcx, M: Machine<'mir, 'tcx>> Memory<'mir, 'tcx, M> { // thing here: one maps to `GlobalAlloc::Static`, this is the "lazy" ID, // and the other one is maps to `GlobalAlloc::Memory`, this is returned by // `const_eval_raw` and it is the "resolved" ID. - // The resolved ID is never used by the interpreted progrma, it is hidden. + // The resolved ID is never used by the interpreted program, it is hidden. + // This is relied upon for soundness of const-patterns; a pointer to the resolved + // ID would "sidestep" the checks that make sure consts do not point to statics! // The `GlobalAlloc::Memory` branch here is still reachable though; when a static // contains a reference to memory that was created during its evaluation (i.e., not // to another static), those inner references only exist in "resolved" form.